Machine learning-based prediction of postoperative nausea and vomiting after spinal anesthesia: A retrospective observational study

作者：Hiroshi Hoshijima, Tomo Miyazaki, Shinichiro Omachi, Daisuke Konno, Shigekazu Sugino, Masanori Yamauchi, Toshiya Shiga, Kentaro Mizuta · 发表于：PLoS ONE · 年份：2026 · DOI：10.1371/journal.pone.0333162 · 研究领域：Nausea and vomiting management、Anesthesia and Pain Management、Intraoperative Neuromonitoring and Anesthetic Effects

Postoperative nausea and vomiting (PONV) is a frequent and serious complication after surgery. PONV also reduces patient satisfaction with surgery under spinal anesthesia and increases medical costs due to prolonged hospitalization. The purpose of this study is to apply artificial intelligence (AI) machine learning analysis to identify risk factors for PONV in patients undergoing surgery with spinal anesthesia. This retrospective study used artificial intelligence to analyze data of adult patients (aged ≥20 years) who underwent surgery under spinal anesthesia at Tohoku University Hospital from January 1, 2010 to December 31, 2022. To evaluate PONV, patients who experienced nausea and/or vomiting or used antiemetics within 24 hours after surgery were extracted from postoperative medical records. The selected data were analyzed after propensity score matching with patients who did not experience PONV. We created an ensemble model for predicting the probability of PONV using five machine learning algorithms: random forest, gradient boosting machine, k-nearest neighbor, multilayer perceptron, and decision tree. Data were available for 4,574 patients. We performed propensity score matching and selected 538 patients for analysis (269 in the PONV group and 269 in the non-PONV group). The use of postoperative fentanyl was identified as the strongest contributor to PONV, followed by duration of surgery, body mass index (BMI), total urine output, and duration of anesthesia. The identif...